Geometrically Constrained Stenosis Editing in Coronary Angiography via Entropic Optimal Transport
Researchers have introduced the OT-Bridge Editor, a novel method for enhancing coronary angiography (CAG) stenosis detection by generating high-quality synthetic imaging data. Addressing the critical scarcity of diverse clinical data, this approach reframes localized image editing as a constrained entropic optimal transport problem. Unlike traditional diffusion-based methods that often lack pixel-level precision, the OT-Bridge Editor leverages geometric information to steer the generation path, ensuring stronger structural control and preservation. Extensive experiments demonstrate that synthesized angiograms produced by this method significantly improve downstream stenosis detection performance. The model achieved substantial relative gains of 27.8% on the public ARCADE benchmark and 23.0% on a multi-center dataset. These results highlight the potential of synthetic data augmentation to overcome data limitations in medical imaging, thereby improving the precision, generalization, and clinical translation of automated diagnostic tools for cardiovascular diseases.
Wire timeline
Geometrically Constrained Stenosis Editing in Coronary Angiography via Entropic Optimal Transport
Researchers have introduced the OT-Bridge Editor, a novel method for enhancing coronary angiography (CAG) stenosis detection by generating high-quality synthetic imaging data. Addressing the critical scarcity of diverse clinical data, this approach reframes localized image editing as a constrained entropic optimal transport problem. Unlike traditional diffusion-based methods that often lack pixel-level precision, the OT-Bridge Editor leverages geometric information to steer the generation path, ensuring stronger structural control and preservation. Extensive experiments demonstrate that synthesized angiograms produced by this method significantly improve downstream stenosis detection performance. The model achieved substantial relative gains of 27.8% on the public ARCADE benchmark and 23.0% on a multi-center dataset. These results highlight the potential of synthetic data augmentation to overcome data limitations in medical imaging, thereby improving the precision, generalization, and clinical translation of automated diagnostic tools for cardiovascular diseases.
cs.AI updates on arXiv.org